"cannot convert parameter from 'short **' to 'void **'"

[Igor Okulist]

int func(void**);

{
short* p = NULL;
func(&p); //<<< here
}

Could somebody remind me why is this not allowed ?
error message: "cannot convert parameter from 'short **' to 'void **'"

Thanks,
Igor

[Neil Cerutti]

On 2005-11-22, Igor Okulist <igor@remove_this_okulist.com> wrote:

int func(void**);

{
short* p = NULL;
func(&p); //<<< here
}

Could somebody remind me why is this not allowed ? error
message: "cannot convert parameter from 'short **' to 'void
**'"

You're attempting to automatically convert a pointer to pointer
to short to a pointer to pointer to void.

A pointer to pointer to void hasn't got the special semantics of
a void pointer.

--
Neil Cerutti

[Ron Natalie]

Igor Okulist wrote:

int func(void**);

{
short* p = NULL;
func(&p); //<<< here
}

Could somebody remind me why is this not allowed ?
error message: "cannot convert parameter from 'short **' to 'void **'"

Because while a short* value can be converted to a void*, it's
not the case that they are the same format. short** and void** are
therefore not convertible.

Imagine that you have an architecture where a char* (and hence
void*) are 32 bits byte addressed and a short pointer is word
addressed.

When you do
short s;
short* sp = &s;
void* vp = sp;
the compiler needs to shift the short value to the left to store
it in the void pointer.

However, if you were allowed to do this:
void** vpp = &sp;

How would the compiler know to convert the short* value?

[Igor Okulist]

> You're attempting to automatically convert a pointer to pointer

to short to a pointer to pointer to void.

A pointer to pointer to void hasn't got the special semantics of
a void pointer.

--
Neil Cerutti

Neil,

Is there a reason why the semantics do not apply to void**? or just
no body bothered to specify it? The stride will always be the same (
lets disregard the near/far possibility).

Thanks,
Igor

[Clark S. Cox III]

On 2005-11-22 13:51:29 -0500, "Igor Okulist"
<igor@remove_this_okulist.com> said:

You're attempting to automatically convert a pointer to pointer
to short to a pointer to pointer to void.

A pointer to pointer to void hasn't got the special semantics of
a void pointer.

--
Neil Cerutti

Neil,

Is there a reason why the semantics do not apply to void**? or just
no body bothered to specify it? The stride will always be the same

Not always. Consider what would happen if sizeof(short*) !=
sizeof(void*), or if the two don't share the same internal
representation (both are possible).

--
Clark S. Cox, III
cl*******@gmail.com

[roberts.noah]

Igor Okulist wrote:

int func(void**);

{
short* p = NULL;
func(&p); //<<< here
}

Could somebody remind me why is this not allowed ?

http://www.parashift.com/c++-faq-lit....html#faq-21.2

ALthough it speaks to classes the same logic would apply.

[roberts.noah]

Igor Okulist wrote:

int func(void**);

{
short* p = NULL;
func(&p); //<<< here
}

Could somebody remind me why is this not allowed ?

http://www.parashift.com/c++-faq-lit....html#faq-21.2

ALthough it speaks to classes the same logic would apply.

[Igor Okulist]

> Not always. Consider what would happen if sizeof(short*) != sizeof(void*),

or if the two don't share the same internal representation (both are
possible).

--
Clark S. Cox, III
cl*******@gmail.com

I'll try to collect all responses here:

Consider what would happen if sizeof(short*) != sizeof(void*)

Well that would make life very complicated. Does that actually happen
(except the near/far pointers)? (this also address Ron's response)

Robert pointed to this faq:
http://www.parashift.com/c++-faq-lit....html#faq-21.2

I disagree (not with the FAQ), but the analogy does not apply here. When
converting
to void* pointer one explicitly casts it to (for lack of better word)
"void". Meaning
that if one is going to dereference it he/she will have to infer the type by
some
other means.

So, yes "Derived** ? Base** " is an error, but "something** -> void**"
still seems to make sense, since void** is not "is-a" relationship but just
a pointer
to a pointer to a memory location... or am I missing some improant step?

Thanks,
Igor

[Clark S. Cox III]

On 2005-11-22 19:16:07 -0500, "Igor Okulist"
<igor@remove_this_okulist.com> said:

Not always. Consider what would happen if sizeof(short*) !=
sizeof(void*), or if the two don't share the same internal
representation (both are possible).

Well that would make life very complicated. Does that actually happen
(except the near/far pointers)?

Yes. There is no requirement that all pointer types have the same size
and representation, and there are many platforms where this happens.

--
Clark S. Cox, III
cl*******@gmail.com

[Neil Cerutti]

On 2005-11-23, Clark S Cox III <cl*******@gmail.com> wrote:

On 2005-11-22 19:16:07 -0500, "Igor Okulist"
<igor@remove_this_okulist.com> said:

Not always. Consider what would happen if sizeof(short*) !=
sizeof(void*), or if the two don't share the same internal
representation (both are possible).

Well that would make life very complicated. Does that actually
happen (except the near/far pointers)?

Yes. There is no requirement that all pointer types have the
same size and representation, and there are many platforms
where this happens.

Moreover, C++ doesn't need more than one kind of generic pointer.

void func(void*)
{
short *s;
func(&s);
}

What would the concept, "generic pointer to generic pointer"
mean, and how would it be used? What would it offer that a void*
doesn't?

--
Neil Cerutti